Hi folks
It looks like there's a kernel OOM killer rewrite in progress. Perhaps
there's a chance of finally getting some sane heuristics for shared
memory in place? There's no indication that it's going to happen without
intervention, but there's talk of rewriting the badness() function that
determines what to kill, so there's more hope than usual.
lwn.net has written about the details. Here's a link that'll let you
read the (otherwise subscriber-only) article in the current edition.
http://lwn.net/SubscriberLink/391222/ce3840ccc36afcd1/
( Please support lwn if you like their work - they're an incredible
resource ).
I'm not on lkml, but may brave the shark-infested waters for this. It'd
be better if people with more of a clue and more background on the issue
got involved, though, which is why I'm posting here.
--
Craig Ringer